Transcriptomic Analysis of Cultured Corneal Endothelial Cells as a Validation for Their Use in Cell Replacement Therapy

Abstract

The corneal endothelium plays a primary role in maintaining corneal homeostasis and clarity and must be surgically replaced with allogenic donor corneal endothelium in the event of visually significant dysfunction. However, a worldwide shortage of donor corneal tissue has led to a search for alternative sources of transplantable tissue. Cultured human corneal endothelial cells (HCEnC) have been shown to restore corneal clarity in experimental models of corneal endothelial dysfunction in animal models, but characterization of cultured HCEnC remains incomplete. To this end, we utilized next-generation RNA sequencing technology to compare the transcriptomic profile of ex vivo human corneal endothelial cells (evHCEnC) with that of primary HCEnC (pHCEnC) and HCEnC lines and to determine the utility of cultured and immortalized corneal endothelial cells as models of in vivo corneal endothelium. Multidimensional analyses of the transcriptome data sets demonstrated that primary HCEnC have a closer relationship to evHCEnC than do immortalized HCEnC. Subsequent analyses showed that the majority of the genes specifically expressed in HCEnC (not expressed in ex vivo corneal epithelium or fibroblasts) demonstrated a marked variability of expression in cultured cells compared with evHCEnC. In addition, genes associated with either corneal endothelial cell function or corneal endothelial dystrophies were investigated. Significant differences in gene expression and protein levels were observed in the cultured cells compared with evHCEnC for each of the genes tested except for AGBL1 and LOXHD1, which were not detected by RNA-seq or qPCR. Our transcriptomic analysis suggests that at a molecular level pHCEnC most closely resemble evHCEnC and thus represent the most viable cell culture-based therapeutic option for managing corneal endothelial cell dysfunction. Our findings also suggest that investigators should perform an assessment of the entire transcriptome of cultured HCEnC prior to determination of their potential clinical utility for the management of corneal endothelial cell failure.

Keywords

Corneal endothelium Transcriptome Gene expression Cell therapy Human corneal endothelial cells (HCEnC)

Introduction

The human cornea comprises several anatomically distinct layers. The most posterior layer is the corneal endothelium, a monolayer of endothelial cells that separates the fibrous structure of the cornea from the aqueous humor. The endothelium is a mosaic of mitotically inactive endothelial cells (23,47) that demonstrate hexagonal morphology and strong basal–apical polarity, features conserved in various vertebrate species (8,56). The hexagonal morphology is an optimal cell geometry that maximizes cell packing (i.e., cell density), minimizes light scatter, and is an outcome of the interplay between distinct biomechanical forces and cellular processes (e.g., cell adhesion and proliferation) that may confer the most favorable physical properties to the endothelium (29). Functionally, the endothelium plays an important role in maintaining hydration of the corneal stroma by selective transport of solutes and water (3,20). Together, these characteristics enable the corneal endothelium to maintain the optical clarity of the cornea.

Corneal endothelial dysfunction secondary to inherited disorders of the endothelium is the most common indication for corneal transplantation in the US (12). Several clinically distinct corneal endothelial dystrophies have been described, but all are characterized by varying degrees of reduced endothelial cell density (1,55). The most severe cases demonstrate an inability of the endothelium to compensate for the loss of endothelial cell function, resulting in corneal edema and a significant reduction in vision. Surgical management is then required and is performed via selective replacement of the corneal endothelium with the posterior lamellar keratoplasty techniques Descemet stripping endothelial keratoplasty (DSEK) and Descemet membrane endothelial keratoplasty (DMEK). While it is generally agreed that these techniques provide superior outcomes when compared to full-thickness penetrating keratoplasty (PK) (10,30), a recent study demonstrated poorer endothelial cell survival for DSEK compared to PK (9). Additionally, surgical intervention carries inherent risks, and accelerated postoperative endothelial cell loss compared to normal corneas limits the survival of all corneal transplants. Moreover, there remains a significant worldwide shortage of suitable donor corneal tissue due to a myriad of cultural, social, religious, and legislative barriers to donation. Thus, there is great interest in alternative approaches to the management of corneal endothelial failure that would obviate the need to recover, process, and transplant a donor cornea with good endothelial viability for each individual with endothelial compromise. The isolation and establishment of donor cornea-derived primary human corneal endothelial cell (pHCEnC) cultures offers the potential for endothelial cells from a single donor to be transplanted into multiple recipients, which could significantly increase the number of both primary and repeat corneal transplants worldwide (21,28,39,44).

Prior to the use of cultured HCEnC for transplantation, these cells will need to be shown to possess similar morphologic and functional characteristics as in vivo HCEnC [with ex vivo HCEnC (evHCEnC) being used as a proxy]. Several HCEnC lines have been established with the expectation that their use in cell replacement therapy would be realized (2,17,26,46,50). These cell lines were reported to show prototypical HCEnC morphology and the expression of a limited number of genes believed to be functional biomarkers of corneal endothelium. Similarly, protocols for the isolation and maintenance of pHCEnC cultures have been established, and characterization of these cells (34,37,41,43,59), including their suitability for use in cell replacement therapy (44), has been performed. While several whole-genome expression studies of evHCEnC have been published (4–6,14–16,24), only limited transcriptomic analyses for pHCEnC (6) and HCEnC lines (19) have been reported. Therefore, we performed a comprehensive transcriptomic analysis of several HCEnC lines and pHCEnC to determine the validity of using cultured HCEnC as a model for in vivo endothelial cell gene expression and to determine the suitability of using cultured HCEnC for management of endothelial cell dysfunction.

Materials and Methods

Primary and Cell Line Cultures

Primary cultures of human corneal endothelial cells were isolated from six corneas (from both eyes of three donors) obtained from various eye banks affiliated with the Vision Share consortium of eye banks (Vision Share, Apex, NC, USA) (Table 1) using a technique that we have adapted for ex vivo donor tissue (14,45). Stripped Descemet membrane with attached endothelium was transferred into complete F99 medium [1:1 ratio of Ham's F-12 nutrient mixture and M199 medium (Life Technologies, Carlsbad, CA, USA), 5% (v/v) fetal bovine serum (Atlanta Biologicals, Flowery Branch, GA, USA), 20 μg/ml ascorbic acid (Sigma-Aldrich, St. Louis, MO, USA), 20 μg/ml insulin (Life Technologies), 10 ng/ml basic fibroblast growth factor (bFGF; PeproTech, Inc., Rocky Hill, NJ, USA), 100 U/ml penicillin, and 100 μg/ml streptomycin (Life Technologies)] and allowed to acclimate to the new growth conditions overnight. The endothelial cells were subsequently detached from Descemet membrane with 0.25% trypsin-EDTA (Life Technologies) for 5 min followed by trituration via gentle passing through a 1-ml pipette tip 10 times. Single cell suspensions of pHCEnC and the corneal endothelial cell lines HCEnC-21T, HCEC-12 (German Collection of Microorganisms and Cell Cultures GmbH, Braunschweig, Lower Saxony, Germany), and HCEC-B4G12 (German Collection of Microorganisms and Cell Cultures) were grown on tissue culture-treated, nonpyrogenic polystyrene plastic coated with 40 μg/cm² chondroitin sulfate A (Sigma-Aldrich) and 40 ng/cm² laminin (Sigma-Aldrich) in phosphate-buffered saline (PBS) for 2 h (2,46,52). The pHCEnC, HCEnC-21T, and HCEC-12 were grown in F99 medium, while the HCEC-B4G12 was maintained in human endothelial-SFM (Life Technologies) supplemented with 10 ng/ml bFGF. All cells were maintained at 37°C in 5% CO₂. When the cells achieved an intact and confluent monolayer, they were incubated overnight in serum-free F99 medium containing only ascorbic acid and antibiotics (pHCEnC, HCEnC-21T, and HCEC-12) or in human endothelial-SFM containing antibiotics (HCEC-B4G12). Cultures of pHCEnC were collected for experiments at passage one.

Table 1.

Donor Tissue Information

Age	Cause of Death	Endothelial Cells per mm²	Preservation Medium	Reason Not Suitable for Transplantation	Death to Cooling	Death to Preservation	Death to Isolation^*
Ex vivo endothelium used for RNA-seq
18	Asphyxiation	3,344, 3,509	Optisol GS	Infiltrates in OU	3 h 20 min	4 h 2 min	2 days
62	Coronary thrombosis	2,933	Optisol GS	Radial keratotomy	4 h 23 min	10 h 53 min	4 days
70	Gastrointestinal bleed	2,899	Life4C	Possible refractive sx	2 h 23 min	5 h 48 min	2 days
Ex vivo endothelium used for qPCR
20	Blunt force trauma	3,390	Life4C	Foreign body (glass)	7 h 17 min	12 h 45 min	2 days
32	Pulmonary embolism	2,793, 2,825	Optisol GS	High-risk behavior	2 h 0 min	14 h 19 min	3 days
48	Anoxic encephalopathy	3,067	Optisol GS	Sepsis	2 h 50 min	11 h 37 min	4 days
55	Myocardial infarction	3,049, 3,135	Optisol GS	Keratotomy scars	3 h 20 min	7 h 45 min	3 days
Ex vivo endothelium used for immunoblotting
49	Hemorrhage	2,874, 2,976	Optisol GS	Stromal opacification	3 h 48 min	13 h 13 min	3 days
62	Coronary thrombosis	3,135	Optisol GS	Radial keratotomy	4 h 23 min	10 h 53 min	4 days
70	Gastrointestinal bleed	2,786	Life4C	Possible refractive sx	2 h 23 min	5 h 48 min	2 days
Ex vivo endothelium used for primary HCEnC cultures
9	Anoxia	3,676, 4,000	Optisol GS	N.D. of unknown etiology	N/A	1 h 42 min	2 days
5	Sudden cardiac death	4,310, 4,505	Optisol GS	Transverse myelitis	1 h 42 min	18 h 42 min	3 days
15	Massive head trauma	3,509, 3,559	Optisol GS	Diluted plasma	2 h 35 min	11 h 15 min	2 days

OU, oculus uterque; sx, surgery; N.D., neurodegenerative disease; N/A, not applicable.

Number of days postmortem before RNA was extracted, protein lysates were prepared, or primary cells were cultured.

Total RNA Isolation

Ten corneas from seven donors were obtained from various eye banks affiliated with the Vision Share consortium of eye banks (Table 1). Descemet membrane (with the attached endothelial cells) was stripped from the donor corneas in preparation for RNA isolation. Where applicable, donor corneas from a single donor were combined into a single tube. Homogenization and total RNA isolation from evHCEnC and cultured cells were performed using TRI Reagent (Life Technologies). TRI Reagent RNA preparations were subsequently purified with the RNeasy Clean-Up Kit (Qiagen, Valencia, CA, USA). The integrity of the isolated RNA was analyzed using the Agilent 2100 Electrophoresis Bioanalyzer System (Agilent Technologies, Inc., Santa Clara, CA, USA), with RNA integrity numbers of 8.1, 8.6, and 9.1 obtained for the three evHCEnC samples.

Next-Generation RNA Sequencing (RNA-seq)

Total RNA was submitted to the UCLA Clinical Microarray Core for processing and sequencing. Briefly, enrichment for poly(A) RNAs was performed using the NEBNext Poly(A) mRNA Magnetic Isolation Module (New England BioLabs, Inc., Ipswich, MA, USA), followed by library preparation using the PrepX Complete ILM DNA Library Kit (WaferGen Biosystems, Inc., Fremont, CA, USA). High-throughput sequencing was performed on the Illumina Hi-seq 2500 (Illumina, Inc., San Diego, CA, USA). Single-end 50-bp reads were obtained and compiled in FASTQ files. A mean of 12,330,531 reads per sample (range 9,497,109–15,392,285 reads) was obtained, which is sufficient for gene expression analysis (31). The FASTQ files and quantitative results are available from the GEO DataSets database [accession number GSE65991; National Center for Biotechnology Information (NCBI), Bethesda, MD, USA].

Next-Generation Sequencing Data Analyses

The FASTQ files containing the RNA-sequencing data were uploaded to the Partek Flow servers (Partek Incorporated, Chesterfield, MO, USA) for alignment to the hg38 genome using the TopHat aligner, which output the results in BAM files. The BAM files were uploaded to the Partek Genomics Suite software, and the reads and read-depth were transformed to reads per kilobase per million (RPKM) values, a normalized quantity that accounts for gene size. These data were annotated using the Ensemble 77 transcript database. Principle component analysis (PCA) and hierarchical clustering were performed in the Partek Genomics Suite software. The gene expression threshold level for positive detection of a transcript was set to a value of 1 RPKM. Differential expression analysis was performed and p values calculated with a one-way ANOVA model using method of moments. Genes expressed specifically in evHCEnC were identified by their absence in ex vivo human corneal epithelial cells (evHCEpC) and ex vivo human corneal fibroblasts cells (evHCFC) and meeting the following statistical criteria: fold-change ≥ 5 and a false discovery rate (step-up) adjusted value of p ≤ 0.05.

Quantitative Polymerase Chain Reaction

Quantitative polymerase chain reaction (qPCR) was used to validate the level of expression as determined by RNA-seq for genes associated with corneal endothelial cell function (AQP1, ATP1A1, CD200, CDH2, GPC4, and ZO1) and the corneal endothelial dystrophies (AGBL1, COL8A2, LOXHD1, SLC4A11, TCF4, and ZEB1). Complementary DNA (cDNA) was synthesized from 100 ng of total RNA isolated from evHCEnC, pHCEnC, and the HCEnC-21T, HCEC-12, and HCEC-B4G12 cell lines using the SuperScript III First-Strand kit (Life Technologies). Subsequently, qPCR was performed on the LightCycler 480 System (Roche, Basel, Switzerland) using the KAPA SYBR FAST qPCR Kit (Kapa Biosystems, Wilmington, MA, USA) and transcript-specific oligonucleotide primers that were obtained from the Harvard Primer Bank database (http://pga.mgh.harvard.edu/primerbank/index.html) (Table 2) (48,49,54). Reaction conditions were as previously described (27). Relative expression was obtained by comparison to the housekeeping gene RAB7A and calculated using the comparative Ct (2^–ΔΔCt) method (32). Relative expression levels were plotted as 2^–ΔCt values.

Table 2.

Quantitative PCR Oligonucleotides

Gene	Primer Bank ID	Forward (5′-3′)	Reverse (5′-3′)
Genes reported as functional markers of corneal endothelium
AQP1	297307115c1	CTGGGCATCGAGATCATCGG	ATCCCACAGCCAGTGTAGTCA
ATP1A1	237681108c2	CTGTGGATTGGAGCGATTCTT	TTACAACGGCTGATAGCACCA
CD200	90903244c3	ACGTCTGTTACCAGCATCCTC	CTTAAAGTCGGTCACAGTCCC
CDH2	215422305c1	TCAGGCGTCTGTAGAGGCTT	ATGCACATCCTTCGATAAGACTG
GPC4	21614524c3	TCCCTCGCAAATTGAAGCTCC	GCAACCGCTAAGCCTTGAG
ZO1	116875766c1	CAACATACAGTGACGCTTCACA	CACTATTGACGTTTCCCCACTC
Genes associated with corneal endothelial dystrophies
AGBL1	120444923c1	CCCTGGTAGACACAGCGAG	GGCCTTCCGTCCAATCTTTTTA
COL8A2	281182524c3	GCTGGCTTAGGCAAACCTG	GGACTCCCACACCGTCTACT
LOXHD1	290542347c1	TGACCCTCTACGGCATCAAC	AGCAATGTCTAGGATCTCCATGA
SLC4A11	291490691c2	TCAATGTCAACCTTGAGATGCAG	TCAGGTACTTTCGGGAGGTGT
TCF4	342672058c3	TGCAAAGCCGAATTGAAGATCG	AGAAGGTCCAATGATTCCATGC
ZEB1	291575187c2	TTACACCTTTGCATACAGAACCC	TTTACGATTACACCCAGACTGC
Housekeeping gene
RAB7A	40807361c1	GTGTTGCTGAAGGTTATCATCCT	GCTCCTATTGTGGCTTTGTACTG

Western Blotting

Four corneas from three donors were used for Western blotting (Table 1). Protein lysates from the five different HCEnC sources were prepared by homogenizing tissue in radioimmunoprecipitation assay (RIPA) buffer [100 mM Tris pH 7.6 (Sigma-Aldrich), 150 mM NaCl (Sigma-Aldrich), 1 mM EDTA (Sigma-Aldrich), 1% deoxycholic acid (Sigma-Aldrich), 1% Triton X-100 (Sigma-Aldrich), 0.1% SDS (Sigma-Aldrich)] and supplemented with fresh 20 mM phenylmethylsulfonyl fluoride (PMSF), 50 mM sodium fluoride (NaF), and protease and phosphatase inhibitors (Life Technologies). A total of 5 μg of whole cell lysate was resolved on a precast NuPAGE Novex 4–12% gradient gel (Life Technologies) by electrophoresis at 40 mA per gel. Following overnight electrotransfer to Immobilon-P (Millipore, Billerica, MA, USA) polyvinylidene fluoride (PVDF) membranes, the membranes were then blocked with 5% milk in TBS-T [100 mM Tris-HCl (Sigma-Aldrich), pH 7.5, 90 g/L NaCl (Sigma-Aldrich), and 1% Tween 20 (Sigma-Aldrich)] for 1 h at room temperature (RT). Incubation with primary antibodies (Table 3) was performed overnight at 4°C in 0.1% milk in TBS-T followed by three washes in TBS-T, then 1-h incubation at RT with peroxidase-coupled secondary antibody. The immunocomplex was detected using Luminata Forte Western HRP Substrate (Millipore) and visualized on Amersham Hyperfilm ECL (GE Healthcare Bio-Sciences Corp., Piscataway, NJ, USA). Detection of the RAB7 protein, a housekeeping gene that regulates vesicular transport, was used as a loading control (11).

Table 3.

Antibodies Used for Immunoblotting

Target Protein	Isotype	Immunogen Species	Clonality	Dilution	Vendor	Catalog Number
Primary antibodies
AQP1	rabbit IgG	human	polyclonal	1:1,000	Millipore	AB3272
ATP1A1	rabbit IgG	human	polyclonal	1:1,000	Cell Signaling Technology	3010
CD200	goat IgG	human	polyclonal	1:1,000	R&D Systems	AF2724
CDH2	rabbit IgG	human	polyclonal	1:1,000	Cell Signaling Technology	4061
COL8A2	rabbit IgG	human	polyclonal	1:1,000	Santa Cruz Biotechnology	sc-82843
GPC4	mouse IgG1	human	monoclonal	1:500	Novus Biologicals	NBP1-45286
RAB7	rabbit IgG	human	monoclonal	1:1,000	Cell Signaling Technology	9367
SLC4A11	rabbit IgG	human	polyclonal	1:1,000	Abcam	Ab93109
TCF4	rabbit IgG	human	monoclonal	1:400	Cell Signaling Technology	2569
ZEB1	rabbit IgG	human	monoclonal	1:500	Cell Signaling Technology	3396
ZO1	rabbit IgG	human	polyclonal	1:1,000	Cell Signaling Technology	5406
Secondary horseradish peroxidase antibodies
Rb-IgG	goat IgG	rabbit	polyclonal	1:30,000	Millipore	AP132P
Ms-IgG	goat IgG	mouse	polyclonal	1:30,000	Millipore	AP124P
Gt-IgG	donkey IgG	goat	polyclonal	1:30,000	Millipore	AB324P

Statistical Analyses

The mean and standard error of the mean (SEM) were graphed for each of the transcript abundance values determined by RNA-seq (RPKM) and qPCR (2^–ΔCt). Statistical testing was performed using one-way ANOVA followed by a post hoc Dunnett's multiple comparison test. Dunnett's multiple comparison test was used to detect a significant (p ≤ 0.05) difference in the mean expression level for each gene in the cultured HCEnC groups versus the mean expression level in evHCEnC. All statistical analyses were performed using a minimum of n = 3, unless otherwise stated. GraphPad Prism version 5.0f (GraphPad Inc., La Jolla, CA, USA) for Mac was used for generating graphs and for statistical analysis.

Results

Cultured HCEnC Demonstrate Prototypical Morphology

The cultured HCEnC were imaged by phase-contrast microscopy before being collected for RNA and protein isolation (Fig. 1). While in vivo corneal endothelium is comprised of flat cells with primarily hexagonal morphology, the pHCEnC and cell lines demonstrated fewer hexagonal cells and flat (pHCEnC) or cobblestone (HCEC-12 and HCEC-B4G12) morphology. These results are consistent with published reports for the pHCEnC and the two HCEnC cell lines (2,46,52).

Figure 1.

Morphology of cultured HCEnC visualized using phase-contrast microscopy. (A) Specular microscopic imaging of human corneal endothelium demonstrates a uniform mosaic of hexagonal cells. (B) pHCEnC demonstrated primarily polygonal rather than hexagonal morphology (representative of 11 primary cultures). The cell lines, (C) HCEnC-21T, (D) HCEC-12, and (E) HCEC-B4G12, also demonstrated a polygonal morphology.

Multidimensional Analyses of Transcriptome Data Sets

PCA and hierarchical clustering (HC) were performed on RNA-seq data sets from HCEnC (Fig. 2). Gene expression data obtained by RNA-seq technology demonstrated a strong association among replicates of the same RNA source by PCA. The samples clustered into distinct groups, which is illustrated along the first three principle component axes [PC #1 (26.3%), PC #2 (16.7%), and PC #3 (11.2%)] and represent the majority of the variation (54.1%) in the samples (Fig. 2A). pHCEnC samples showed a close relationship with evHCEnC samples, while the immortalized cells demonstrated a more distant relationship. Similarly, hierarchical clustering of the HCEnC groups also demonstrated a strong association among replicates of the same RNA source (Fig. 2B). Because the HCEC-B4G12 cell line is a clone of the HCEC-12 cell line, they showed a close relationship by both PCA and HC.

Figure 2.

Multidimensional analysis shows distinct clustering of the five HCEnC groups by RNA source. PCA and HC were used to determine the relationships between the transcriptome data sets from the different HCEnC groups. (A) PCA of the transcriptomic data sets demonstrated that samples clustered into distinct groups defined by their RNA source. (B) HC of the HCEnC groups using the transcriptome data sets demonstrated a similar association pattern observed with PCA.

Comparative Analysis of Whole-Genome Gene Expression in evHCEnC Versus Cultured HCEnC

To determine the genes that are shared and unique between evHCEnC and the cultured HCEnC groups, we defined any gene across the genome with a mean RPKM value ≥ 1 as being expressed. These expressed genes were compiled into gene lists for each of the HCEnC groups (data not shown). Subsequently, the gene lists for the cultured HCEnC were each compared separately to the gene list for the evHCEnC (Fig. 3 and Table 4). Compared with evHCEnC, pHCEnC demonstrated the greatest percentage of commonly expressed genes (88.3%; 11,684/13,234), followed by HCEnC-21T (86.3%; 11,424/13,234), HCEC-12 (84.1%; 11,133/13,234), and HCEC-B4G12 (83.5%; 11,053/13,234) (Fig. 3). Analysis of unique genes in the cultured cells showed that in pHCEnC de novo gene expression or the change in total genes expressed was 12.0% (1,583/13,234) compared with the total genes expressed in evHCEnC, while a de novo gene expression of 24.0% (3,172/13,234), 19.8% (2,620/13,234), and 19.4% (2,571/13,234) was observed for HCEnC-21T, HCEC-12, and HCEC-B4G12, respectively (Table 4). In addition, the total number of unique genes (unique to either evHCEnC or to each HCEnC group) is lowest in pHCEnC (21.1%; 3,133/14,817) and about the same (approximately 30%) for each of the HCEnC lines.

Figure 3.

Gene expression comparison between evHCEnC and cultured HCEnC identifies common and distinct gene expression patterns. Genes were compiled in a list for each of the HCEnC groups that met the criterion of being expressed at a level ≥ 1 RPKM. Comparisons were made between the genes expressed in evHCEnC and in each of the cultured HCEnC groups: (A) pHCEnC, (B) HCEnC-21T, (C) HCEC-12, and (D) HCEC-B4G12. Percentage = (number of common genes)/(number of evHCEnC expressed genes) × 100.

Table 4.

Comparative Analysis of Genes Expressed in the HCEnC Groups

HCEnC Group (vs. evHCEnC)	Change in Total Genes Expressed (%)^*	Genes Unique to evHCEnC^†	Genes Unique to HCEnC Group^‡	Total Unique Genes
pHCEnC	+12.0	1,550 (10.5%)	1,583 (10.7%)	3,133 (21.1%)
HCEnC-21T	+24.0	1,810 (11.0%)	3,172 (19.3%)	4,982 (30.4%)
HCEC-12	+19.8	2,101 (13.3%)	2,620 (16.5%)	4,721 (29.8%)
HCEC-B4G12	+19.4	2,181 (13.8%)	2,571 (16.3%)	4,752 (30.1%)

Gene expression is defined as the genes whose transcript levels were at or above 1 RPKM.

Percentage = (genes unique to HCEnC group)/(genes expressed in evHCEnC) × 100.

†

Percentage = (genes unique to evHCEnC)/(genes unique to evHCEnC + total genes expressed in HCEnC group) × 100.

‡

Percentage = (genes unique to HCEnC group)/(genes unique to HCEnC group + total genes expressed in evHCEnC) × 100.

Identification of evHCEnC-Specific Genes by Absence of Expression in evHCEpC and evHCFC

Endothelium-specific gene expression was determined for evHCEnC (Table 5). All three main corneal cell types (epithelial, fibroblast, and endothelial) were isolated from ex vivo cornea tissue, and the transcriptome profile for each cell type was obtained (data not shown). The genes specific to evHCEnC were determined by 1) filtering out genes with an RPKM value below 1; 2) filtering out genes expressed (> 1 RPKM) in evHCEpC and/or evHCFC); and 3) performing differential gene expression analysis under the criteria described above. One hundred thirty-eight genes were identified as specific to evHCEnC, of which 77% (106/138) were protein coding (Table 5).

Table 5.

Expression of 138 evHCEnC-Specific Genes in Cultured HCEnC

Gene Symbol	Gene Type	evHCEnC (RPKM^*)	pHCEnC (RPKM)	HCEnC-21T (RPKM)	HCEC-12 (RPKM)	HCEC-B4G12 (RPKM)
CA3	protein coding	1066	73	0	0	0
RP11-317J10.2	RNA gene	160	13	1	0	0
ETNPPL ^†	protein coding	74	0	0	0	0
CDH2	protein coding	63	149	54	41	27
PIP5K1B	protein coding	52	74	0	0	1
LRRN1	protein coding	48	54	0	0	0
ADCYAP1R1 ^†	protein coding	46	0	0	0	0
MRGPRX3 ^†	protein coding	45	0	0	0	0
RPL39P40	pseudogene	43	16	5	0	0
SCDP1	pseudogene	43	23	19	10	18
POU6F2	protein coding	36	25	0	0	0
RPRML¶	protein coding	35	0	0	0	0
RP11-149I23.3	novel antisense	35	1	0	0	0
C1QL1	protein coding	33	11	8	20	33
NELL2	protein coding	33	0	1	0	0
OGDHL	protein coding	31	6	5	0	0
PVALB ^†	protein coding	30	0	0	0	0
AJAP1	protein coding	26	6	0	0	0
MIR3652	RNA gene	25	18	0	0	0
COL4A6	protein coding	25	17	2	4	1
PLCE1	protein coding	25	2	4	5	3
PLCE1-AS1	RNA gene	24	33	0	0	18
CTD-2303H24.2	novel transcript	24	3	2	3	3
RP4-794I6.4	lincRNA	24	4	0	0	0
FGF9	protein coding	23	10	0	0	0
RNU2-1 ^†	RNA gene	21	0	0	0	0
KISS1	protein coding	20	4	0	11	1
GNG4	protein coding	20	0	5	2	0
MCHR1 ^†	protein coding	20	0	0	0	0
MRAP2	protein coding	19	7	2	0	0
RP11-92B11.3	novel antisense	19	1	1	0	0
XAGE2B	protein coding	19	7	0	0	0
SOWAHA	protein coding	19	10	2	0	0
SLC9A7	protein coding	19	16	6	9	6
TMEM130	protein coding	18	18	0	0	0
GAL3ST3	protein coding	18	15	0	0	0
ITPR1	protein coding	18	7	0	2	0
AC055723.1	miRNA	17	39	0	2	3
PLK5 ^†	protein coding	17	0	0	0	0
WNT8B	protein coding	17	4	0	0	0
SHC4	protein coding	16	10	0	0	0
NRXN3	protein coding	16	1	0	0	0
WISP1 ^†	protein coding	15	0	0	0	0
CCDC144B	pseudogene	15	2	1	3	3
CHGB	protein coding	15	3	0	0	0
FRMPD1	protein coding	15	1	0	0	0
CLDN10	protein coding	14	0	22	0	0
GLIS3	protein coding	14	11	0	1	0
RGS7BP ^†	protein coding	13	0	0	0	0
MAPK4	protein coding	13	2	0	0	0
SYT13	protein coding	13	1	0	0	0
CA9	protein coding	13	0	0	36	16
RP11-175B9.3	pseudogene	13	2	3	0	0
RBP1	protein coding	12	4	7	0	0
PLEKHH1	protein coding	12	5	3	3	2
PPARGC1B	protein coding	11	2	0	0	0
RGS9	protein coding	11	6	2	1	1
CCDC144A	protein coding	11	1	4	0	0
LINC01497 ^†	lincRNA	11	0	0	0	0
CRB2	protein coding	10	13	0	0	0
FAM106A	protein coding	10	2	0	2	2
PLCH1	protein coding	10	1	1	0	0
RP11-104H15.7	novel transcript	10	2	0	0	0
RNF182	protein coding	10	3	0	1	0
TMEM160	protein coding	10	6	4	6	6
CCDC144CP ^†	pseudogene	9	0	0	0	0
RP11-321G12.1	RNA gene	9	2	0	0	0
TAC3	protein coding	9	1	0	0	0
PPP1R1B ^†	protein coding	9	0	0	0	0
GOLT1A ^†	protein coding	9	0	0	0	0
SEMA6D	protein coding	9	2	3	0	0
NPNT	protein coding	9	4	12	0	0
ALDH1A2	protein coding	9	0	20	0	0
SORCS3	protein coding	9	1	0	0	0
SALL2	protein coding	8	7	19	9	8
GLIS1	protein coding	8	2	1	0	2
U12 ^†	snRNA	8	0	0	0	0
RANBP3L ^†	protein coding	8	0	0	0	0
AC144652.1	lincRNA	8	0	2	1	3
SPOCK2	protein coding	8	7	0	0	0
COX6B2 ^†	protein coding	8	0	0	0	0
CEND1 ^†	protein coding	8	0	0	0	0
CA14	protein coding	8	2	0	0	0
DMKN	protein coding	8	2	4	1	4
GBGT1	protein coding	7	4	0	0	0
RNU4ATAC	RNA gene	7	0	0	0	2
OR51E2	protein coding	7	4	0	0	0
C2orf72	protein coding	7	0	2	2	2
STC2	protein coding	7	3	1	21	42
FRMD3	protein coding	7	4	0	0	0
KCND3 ^†	protein coding	7	0	0	0	0
KIF26B	protein coding	7	7	0	0	0
RNU5E-4P ^†	pseudogene	7	0	0	0	0
ZDHHC8P1 ^†	pseudogene	7	0	0	0	0
GPR158 ^†	protein coding	7	0	0	0	0
PNMAL1	protein coding	7	7	9	2	6
NEFH ^†	protein coding	7	0	0	0	0
PIFO ^†	protein coding	7	0	0	0	0
MED12L	protein coding	7	1	1	3	3
RP11-632K5.2	sense intronic	7	8	0	0	0
CCDC8	protein coding	7	5	13	0	0
ENPP4	protein coding	7	1	8	6	5
SNORD3C ^†	RNA gene	6	0	0	0	0
CA10 ^†	protein coding	6	0	0	0	0
SMOC1	protein coding	6	2	5	3	0
ADORA1	protein coding	6	1	0	1	2
RP11-380D23.2	lincRNA	6	4	2	0	2
LIX1	protein coding	6	4	0	0	0
SNX30	protein coding	6	16	7	7	8
EPHB1	protein coding	6	11	0	0	0
CYFIP2	protein coding	6	5	21	2	4
RGS4	protein coding	6	2	0	1	0
DCC	protein coding	6	1	0	0	0
MLLT3	protein coding	6	8	2	4	3
SPTB	protein coding	6	0	2	0	0
COX4I2 ^†	protein coding	6	0	0	0	0
RP11-118K6.3 ^†	lincRNA	6	0	0	0	0
HYKK	protein coding	6	0	2	1	1
HIF1A-AS2 ^†	RNA gene	6	0	0	0	0
FCGBP ^†	protein coding	6	0	0	0	0
FAM181B	protein coding	5	1	0	0	0
CTD-2653D5.1	novel antisense	5	14	0	0	0
GDF5	protein coding	5	12	0	0	0
PVRL3	protein coding	5	4	4	4	3
SCAI	protein coding	5	1	4	3	3
CDK14	protein coding	5	11	6	1	4
GRIP1	protein coding	5	2	1	0	0
LOXL1-AS1	RNA gene	5	23	3	4	5
CYP1A2	protein coding	5	5	0	0	0
RP11-211N11.5 ^†	novel antisense	5	0	0	0	0
SNCA	protein coding	5	9	4	2	0
MGAT3	protein coding	5	0	7	0	0
AC064875.2	lincRNA	5	1	0	0	0
B9D1	protein coding	5	3	3	5	7
ARHGAP40 ^†	protein coding	5	0	0	0	0
ACSS3 ^†	protein coding	5	0	0	0	0
SCAMP5	protein coding	5	9	5	2	3
RP11-382B18.5 ^†	pseudogene	5	0	0	0	0
Total genes expressed		138	93	53	41	39

Genes (26) expressed in all HCEnC groups are shown in bold.

Genes with a mean RPKM value of ≥ 1 (< 1 in evHCEpC and evHCFC) and a fold change ≥ 5 compared with evHCEpC and evHCFC were considered to be specific to evHCEnC.

†

Genes (34) not expressed in any of the cultured HCEnC groups.

Analysis of evHCEnC-Specific Gene Expression in Cultured HCEnC

The percentage of the evHCEnC-specific genes expressed in each of the cultured HCEnC groups was determined by gene list comparison and hierarchical clustering analysis (Fig. 4). Comparison of the 138 evHCEnC-specific genes with the expressed genes (> 1 RPKM) in each of the cultured cells demonstrated that 67% (93) of the 138 evHCEnC-specific genes were expressed in pHCEnC, 38% (53/138) were expressed in HCEnC-21T, 30% (41/138) were expressed in HCEC-12, and 28% (39/138) were expressed in HCEC-B4G12 (Fig. 4A and Table 5). Nineteen percent (26/138) of the evHCEnC-specific genes were expressed in all of the cultured HCEnC groups, while 25% (34/138) were not expressed in any of the cultured HCEnC groups (Fig. 4A and Table 5).

Figure 4.

Expression of the 138 evHCEnC-specific genes in cultured HCEnC. (A) Comparison of the genes specific to evHCEnC with the expressed genes in four different cultured HCEnC groups. Twenty-six genes (large bold font, center) were common to all five HCEnC groups, and 34 (large bold font, lower right) genes were only expressed in the evHCEnC group (Table 5). The numbers that are of medium bolded font represent those expressed in evHCEnC and at least one of the cultured HCEnC groups. The numbers that are small and not bold are those that are expressed in the corresponding cultured HCEnC group(s) but are otherwise not a focus of the current study. (B) Hierarchical clustering of the HCEnC groups using the 138 evHCEnC-specific genes demonstrated distinct expression profiles in cultured HCEnC.

HC analysis of the expression profile of the 138 evHCEnC-specific genes compared to the expression profile of these genes in primary and immortalized cells was performed as well (Fig. 4B). pHCEnC demonstrated an expression profile that was more closely related to evHCEnC than those of the cell lines, reflecting the results obtained by gene list comparison (Fig. 4B). Of the immortalized cells, HCEnC-21T demonstrated an expression profile that was more closely related to evHCEnC than the HCEC-12 and HCEC-B4G12 (Fig. 4B).

Evaluation of the Expression of Genes Associated with HCEnC Function or Disease

Genes Associated with Corneal Endothelial Cell Function

The transcription of six genes that are considered to be functional markers of HCEnC (AQP1, ATP1A1, CD200, CDH2, GPC4, and ZO1) was initially assessed by RNA-seq (Fig. 5), and subsequently validated by qPCR (Fig. 5) and Western blotting (Fig. 6A). In evHCEnC, the expression of AQP1, ATP1A1, CDH2, GPC4, and ZO1 was detected above the threshold of 1 RPKM by RNA-seq, with CD200 being detected below this threshold. While the expression of all six genes was detected by qPCR, CD200 was detectable only at relatively high cycles (~33 cycles), suggesting weak or no expression. Western blotting demonstrated detectable levels of AQP1, ATP1A1, CDH2, GPC4, and ZO1 but failed to detect CD200. In pHCEnC, transcripts for all of the functional marker genes were detected by RNA-seq, with expression levels of CD200, CDH2, and GPC4 significantly greater and ATP1A1 significantly less than that in evHCEnC. Quantitative PCR confirmed statistically significant differential expression for ATP1A1, CD200, and GPC4, but not for CDH2. Western blotting demonstrated detectable levels of all six encoded proteins. In HCEnC-21T, the expression of ATP1A1, CDH2, GPC4, and ZO1 was detected above the threshold of 1 RPKM by RNA-seq, with CD200 and AQP1 being detected below this threshold. Only ATP1A1 showed a significant difference in expression compared with that in evHCEnC, which was confirmed by qPCR. Although the expression of AQP1 was approximately 1,000-fold less than in evHCEnC, the difference was not statistically significant (although it was by qPCR). In HCEC-12 and HCEC-B4G12, the expression of ATP1A1, CDH2, and ZO1 was detected by RNA-seq, while GPC4 demonstrated detectable levels (> 1 RPKM) only in HCEC-B4G12, and AQP1 and CD200 were not detected in either cell line. The expression levels of ATP1A1 were significantly lower and ZO1 significantly greater than in evHCEnC for both HCEC-12 and HCEC-B4G12, while the expression level for CDH2 was significantly lower in HCEC-B4G12 only. Quantitative PCR confirmed significant differential expression for ATP1A1 in both HCEC-12 and HCEC-B4G12 and CDH2 in HCEC-B4G12 but not ZO1 in either cell line. Western blotting demonstrated detectable levels of ATP1A1, CDH2, and ZO1 in each cell line, but no detection of AQP1, CD200, or GPC4 in any of the three cell lines.

Figure 5.

Transcript abundance of genes associated with HCEnC function was measured using RNA-seq and validated by qPCR. The dashed line in the RNA-seq graph for CD200 denotes the RPKM threshold (RPKM≥ 1) above which a gene is considered expressed. The evHCEnC are represented as black bars (n = 4), the pHCEnC are represented as white bars (n = 3), and the three cell lines are represented as gray bars (n = 3). Error bars represent mean ± SEM of RPKM (RNA-seq) or 2^–ΔCt (qPCR) values; *p≤0.05, **p≤0.01, ***p≤0.001.

Figure 6.

Proteins encoded by the genes associated with HCEnC function and corneal endothelial dystrophies were assessed by Western blotting. (A) Proteins from 5/6 (AQP1, ATP1A1, CDH2, and ZO1), 6/6 (AQP1, ATP1A1, CDH2, GPC4, and ZO1), and 3/6 (ATP1A1, CDH2, and ZO1) genes associated with HCEnC function were detected in evHCEnC, pHCEnC, and the three HCEnC lines, respectively. (B) Proteins from 4/4 (COL8A2, SLC4A11, TCF4, and ZEB1), 3/4 (COL8A2, SLC4A11, and ZEB1), 1/4 (ZEB1), and 2/4 (COL8A2 and ZEB1) genes associated with corneal endothelial dystrophies were detected in evHCEnC, pHCEnC, HCEnC-21T, and the two (HCEC-12 and HCEC-B4G12) related cell lines, respectively. These are representative blots from three independent samples for each HCEnC group (Table 1). The detection of the RAB7 protein was used as a loading control.

Genes Associated with Corneal Endothelial Dystrophies

The transcription of genes associated with corneal endothelial dystrophies (AGBL1, COL8A2, LOXHD1, SLC4A11, TCF4, and ZEB1) was assessed by RNA-seq (Fig. 7) and the levels validated by qPCR (Fig. 7) and Western blotting (Fig. 6B). In evHCEnC, the expression of four of the genes (COL8A2, SLC4A11, TCF4, and ZEB1) was detected by RNA-seq above the 1-RPKM threshold. Expression of each of the four genes was confirmed with qPCR and Western blotting. The AGBL1 transcript was not detected by either RNA-seq or qPCR, while transcript levels for LOXHD1 were only detected at an average cycle of ~38 in two of the three samples. As neither gene was detected by qPCR, Western blotting for the encoded proteins was not performed. In pHCEnC, transcripts for COL8A2, SLC4A11, TCF4, and ZEB1 were detected by RNA-seq, with expression levels of TCF4 being significantly greater and SLC4A11 significantly less than in evHCEnC. While these results were confirmed with qPCR, Western blotting failed to detect TCF4. In HCEnC-21T, HCEC-12, and HCEC-B4G12, RNA-seq demonstrated expression for SLC4A11 and ZEB1 in all three cell lines, COL8A2 in HCEC-12 and HCEC-B4G12, and TCF4 expression in HCEnC-21T and HCEC-12. AGBL1 and LOXHD1 were not expressed in any of the three cell lines. The expression levels of COL8A2, SLC4A11, and TCF4 were significantly lower and ZEB1 significantly greater in all three cell lines compared with evHCEnC, which was confirmed by qPCR. Western blotting demonstrated detectable levels of COL8A2 in HCEC-12 and HCEC-B4G12 and ZEB1 in all three cell lines.

Figure 7.

The transcript abundance of genes associated with corneal endothelial dystrophies was measured using RNA-seq and validated by qPCR. The dashed line in the RNA-seq graph for AGBL1, LOXHD1, and ZEB1 denotes the RPKM threshold (RPKM≥1) above which a gene is considered expressed. Statistical testing was performed using one-way ANOVA followed by a post hoc Dunnett's multiple comparison test. The evHCEnC are shown as black bars (n = 4), the pHCEnC are represented as white bars (n = 3), and the three cell lines are represented as gray bars (n = 3). Error bars represent mean ± SEM of RPKM (RNA-seq) or 2^–ΔCt (qPCR) values; *p≤0.05, **p≤0.01, ***p≤0.001.

Discussion

The investigation of the functional mechanisms responsible for normal and pathological corneal endothelial cell function remains hampered by the lack of available robust techniques in culturing corneal endothelial cells. The recent establishment of corneal endothelial cell lines using the SV40 TAg (HCEC-12 and HCEC-B4G12) and telomerase expression (HCEnC-21T) for immortalization has presented a unique opportunity for investigators to perform biologically significant experiments in easy-to-maintain cell lines (2,46,52). The initial characterization of these cell lines was performed by assessing the expression of genes considered markers of endothelial cell function, such as AQP1, ATP1A1, CDH2, and ZO1. However, we demonstrate significant differences in the expression of these genes and the proteins they encode in pHCEnC and HCEnC lines compared to their expression in evHCEnC. Thus, while cell lines remain a valuable tool in the absence of robust primary cell culturing techniques (7), investigators should be aware of these differences and plan experiments using these cell lines accordingly.

More recently, GPC4 and CD200 were reported as distinctly expressed only in the corneal endothelium (5), leading to interest in their use as corneal endothelial cell markers. However, we found that CD200 and its encoded protein were not detected in evHCEnC or any of the cell lines but was detected in pHCEnC. In addition, GPC4 expression in evHCEnC was significantly lower than in pHCEnC by both RNA-seq and qPCR, with no GPC4 detected by Western blot in the cell lines. There are myriad reasons for these differences, but they may be due to differing experimental protocols, including the age of the cornea donor, tissue processing techniques, and the method utilized to detect expression. Because of these conflicting results, further investigation is warranted to determine whether CD200 is in fact expressed in the corneal endothelium. Similarly, the genes implicated as playing a pathogenic role in the corneal endothelial dystrophies demonstrated significant differences in expression between the cultured and evHCEnC. Subsequent characterization of pHCEnC and endothelial cell lines will involve determining whether these cells possess any aberrant characteristics that could complicate elucidation of the molecular basis of HCEnC function and dysfunction. For example, an investigator interested in the role of AQP1 in corneal endothelial cell function would be unable to utilize HCEC-12, HCEC-B4G12, and HCEnC-21T given the absence of AQP1. Thus, it is not always possible to extrapolate from what is observed in cultured endothelial cells what would be expected in vivo, which has important implications in terms of the research utility and therapeutic potential of cultured HCEnC.

Our results indicate that the transcriptome of pHCEnC is more similar to that of evHCEnC (and thus presumably to the transcriptome of in vivo HCEnC) than are the transcriptomes of the endothelial cell lines. However, use of pHCEnC for research and for potential clinical applications is restrained by phenotypic alterations and limited growth using commonly employed cell culturing techniques. Identifying the growth conditions that support the growth of cells with a transcriptome profile similar to evHCEnC will provide researchers with a more accurate model of in vivo HCEnC and a potential source of endothelial cells for management of endothelial dysfunction. Optimization of HCEnC culturing techniques should take into consideration several anatomic and physiological features of in vivo HCEnC (60), such as 1) adherence to a complex milieu of extracellular matrix proteins (13,18,22,38,40,51,57), 2) contact with physiological proteins and other factors present in aqueous humor (25,33,36,42,53), 3) exposure to appropriate biomechanical forces (35,40,57), and 4) maintenance of a confluent semipermeable layer of cells with strong apicobasal polarity. Replication of each of these features in a single culturing method would be challenging but would represent a significant advance in the development of cultured HCEnC that closely resemble in vivo HCEnC.

In addition, our findings caution researchers against reaching the conclusion that the expression of a few functional marker genes may be relied upon to characterize cultured HCEnC. This point is particularly salient given the recent report by Chng et al. suggesting that a panel of three genes (COL8A2, SLC4A11, and CYYR1) is sufficient to ascertain the clinical viability of stem cell-derived HCEnC (6). They identified CYYR1 as the most robust marker for ascertaining the HCEnC type, but our analysis did not confirm this since we observed CYYR1 expression in evHCFC at an average level of 2.4 RPKM (above our 1 RPKM expression threshold) as well as evHCEnC (13.6 RPKM). A more sound expression-based assessment may include each of the 106 protein-coding genes specific to evHCEnC (Table 5) or the 22 protein-coding genes specific to evHCEnC that are expressed in pHCEnC and each of the three endothelial cells lines tested. A more recent study identified five genes (CLRN1, MRGPRX3, HTR1D, GRIP1, and ZP4) that demonstrated high HCEnC specificity, but our analyses identified only MRGPRX3 and GRIP1 as expressed only in evHCEnC (58).

We have identified a set of evHCEnC-specific genes that can be utilized in a panel for ascertaining the nature of cultured HCEnC. We expect that some of these genes may be removed from the panel due to inconsistent results being obtained in future studies. However, we believe that the use of such a panel is essential to more thoroughly characterize cultured HCEnC than is possible using only a few functional markers of HCEnC. An expression panel including the 106 protein-coding genes specific to evHCEnC represents only about 0.5% of known protein-coding genes and thus represents a reasonable balance between being both comprehensive and specific. Although a single (or several) biomarker is informative and valuable in, for example, targeted gene therapy or antibody-mediated purification of HCEnC from a mixed population of cells (i.e., positive selection), we do not believe that this number of biomarkers is sufficient to accurately characterize cultured HCEnC for use in cell transplantation therapy. Thus, the identification of these HCEnC-specific genes is an important step in the characterization of cultured HCEnC and validation of their experimental and potential clinical utility.

Footnotes

Acknowledgments

We acknowledge Dr. Ula V. Jurkunas for kindly providing us with the HCEnC-21T. Support provided by National Eye Institute grants R01 EY022082 (A.J.A.) and P30 EY000331 (core grant), the Walton Li Chair in Cornea and Uveitis and an unrestricted grant from Research to Prevent Blindness. The authors declare no conflicts of interest.

References

Aldave

A. J.

; Han

; Frausto

R. F.

Genetics of the corneal endothelial dystrophies: An evidence-based review. Clin. Genet. 84(2): 109–119; 2013.

Bednarz

; Teifel

; Friedl

; Engelmann

Immortalization of human corneal endothelial cells using electroporation protocol optimized for human corneal endothelial and human retinal pigment epithelial cells. Acta Ophthalmol. Scand. 78(2): 130–136; 2000.

Bonanno

J. A.

Identity and regulation of ion transport mechanisms in the corneal endothelium. Prog. Retin. Eye Res. 22(1): 69–94; 2003.

Chen

; Huang

; Nakatsu

M. N.

; Xue

; Deng

S. X.

; Fan

Identification of novel molecular markers through transcriptomic analysis in human fetal and adult corneal endothelial cells. Hum. Mol. Genet. 22(7): 1271–1279; 2013.

Cheong

Y. K.

; Ngoh

Z. X.

; Peh

G. S.

; Ang

H. P.

; Seah

X. Y.

; Chng

; Colman

; Mehta

J. S.

; Sun

Identification of cell surface markers glypican-4 and CD200 that differentiate human corneal endothelium from stromal fibroblasts. Invest. Ophthalmol. Vis. Sci. 54(7): 4538–4547; 2013.

Chng

; Peh

G. S.

; Herath

W. B.

; Cheng

T. Y.

; Ang

H. P.

; Toh

K. P.

; Robson

; Mehta

J. S.

; Colman

High throughput gene expression analysis identifies reliable expression markers of human corneal endothelial cells. PLoS One 8(7): e67546; 2013.

Choi

J. S.

; Kim

E. Y.

; Kim

M. J.

; Khan

F. A.

; Giegengack

; D'Agostino

Jr. ; Criswell

; Khang

; Soker

Factors affecting successful isolation of human corneal endothelial cells for clinical use. Cell Transplant. 23(7): 845–854; 2014.

Collin

S. P.

; Collin

H. B.

A comparative study of the corneal endothelium in vertebrates. Clin. Exp. Optom. 81(6): 245–254; 1998.

Coster

D. J.

; Lowe

M. T.

; Keane

M. C.

; Williams

K. A.

Australian Corneal Graft Registry Contributors. A comparison of lamellar and penetrating keratoplasty outcomes: A registry study. Ophthalmology 121(5): 979–987; 2014.

10.

Dapena

; Ham

; Melles

G. R.

Endothelial keratoplasty: DSEK/DSAEK or DMEK - The thinner the better?

Curr. Opin. Ophthalmol. 20(4): 299–307; 2009.

11.

Eisenberg

; Levanon

E. Y.

Human housekeeping genes, revisited. Trends Genet. 29(10): 569–574; 2013.

12.

Eye Bank Association of America 2013 EBAA Statistical Report. Washington, DC: Eye Bank Association of America; 2014.

13.

Fan

; Zhao

; Ma

; Xu

; Zhao

; Xu

Establishment of a continuous untransfected human corneal endothelial cell line and its biocompatibility to denuded amniotic membrane. Mol. Vis. 17: 469–480; 2011.

14.

Frausto

R. F.

; Wang

; Aldave

A. J.

Transcriptome analysis of the human corneal endothelium. Invest. Ophthalmol. Vis. Sci. 55(12): 7821–7830; 2014.

15.

Gottsch

J. D.

; Bowers

A. L.

; Margulies

E. H.

; Seitzman

G. D.

; Kim

S. W.

; Saha

; Jun

A. S.

; Stark

W. J.

; Liu

S. H.

Serial analysis of gene expression in the corneal endothelium of Fuchs' dystrophy. Invest. Ophthalmol. Vis. Sci. 44(2): 594–599; 2003.

16.

Gottsch

J. D.

; Seitzman

G. D.

; Margulies

E. H.

; Bowers

A. L.

; Michels

A. J.

; Saha

; Jun

A. S.

; Stark

W. J.

; Liu

S. H.

Gene expression in donor corneal endothelium. Arch. Ophthalmol. 121(2): 252–258; 2003.

17.

Griffith

; Osborne

; Munger

; Xiong

; Doillon

C. J.

; Laycock

N. L.

; Hakim

; Song

; Watsky

M. A.

Functional human corneal equivalents constructed from cell lines. Science 286(5447): 2169–2172; 1999.

18.

Gruschwitz

; Friedrichs

; Valtink

; Franz

C. M.

; Muller

D. J.

; Funk

R. H.

; Engelmann

Alignment and cell-matrix interactions of human corneal endothelial cells on nanostructured collagen type I matrices. Invest. Ophthalmol. Vis. Sci. 51(12): 6303–6310; 2010.

19.

Ha Thi

B. M.

; Campolmi

; He

; Pipparelli

; Manissolle

; Thuret

J. Y.

; Piselli

; Forest

; Peoc'h

; Garraud

; Gain

; Thuret

Microarray analysis of cell cycle gene expression in adult human corneal endothelial cells. PLoS One 9(4): e94349; 2014.

20.

Harris

J. E.

Current thoughts on the maintenance of corneal hydration in vivo. Arch. Ophthalmol. 78(2): 126–132; 1967.

21.

Hatanaka

; Koizumi

; Okumura

; Takahashi

; Tanioka

; Young

R. D.

; Jones

F. E.

; Quantock

A. J.

; Kinoshita

A study of host corneal endothelial cells after non-descemet stripping automated endothelial keratoplasty. Cornea 32(1): 76–80; 2013.

22.

Hsieh

; Baum

Effects of fibroblastic and endothelial extracellular matrices on corneal endothelial cells. Invest. Ophthalmol. Vis. Sci. 26(4): 457–463; 1985.

23.

Joyce

N. C.

; Navon

S. E.

; Roy

; Zieske

J. D.

Expression of cell cycle-associated proteins in human and rabbit corneal endothelium in situ. Invest. Ophthalmol. Vis. Sci. 37(8): 1566–1575; 1996.

24.

Jun

A. S.

; Liu

S. H.

; Koo

E. H.

; Do

D. V.

; Stark

W. J.

; Gottsch

J. D.

Microarray analysis of gene expression in human donor corneas. Arch. Ophthalmol. 119(11): 1629–1634; 2001.

25.

Kim

; Kim

J. J.

; Hyon

J. Y.

; Chung

E. S.

; Chung

T. Y.

; Yi

; Wee

W. R.

; Shin

Y. J.

The effects of different culture media on human corneal endothelial cells. Invest. Ophthalmol. Vis. Sci. 55(8): 5099–5108; 2014.

26.

Kim

H. J.

; Ryu

Y. H.

; Ahn

J. I.

; Park

J. K.

; Kim

J. C.

Characterization of immortalized human corneal endothelial cell line using HPV 16 E6/E7 on lyophilized human amniotic membrane. Korean J. Ophthalmol. 20(1): 47–54; 2006.

27.

Kim

M. J.

; Frausto

R. F.

; Rosenwasser

G. O.

; Bui

; Le

D. J.

; Stone

E. M.

; Aldave

A. J.

Posterior amorphous corneal dystrophy is associated with a deletion of small leucine-rich proteoglycans on chromosome 12. PLoS One 9(4): e95037; 2014.

28.

Koizumi

; Okumura

; Kinoshita

Development of new therapeutic modalities for corneal endothelial disease focused on the proliferation of corneal endothelial cells using animal models. Exp. Eye Res. 95(1): 60–67; 2012.

29.

Lecuit

; Lenne

P. F.

Cell surface mechanics and the control of cell shape, tissue patterns and morphogenesis. Nat. Rev. Mol. Cell Biol. 8(8): 633–644; 2007.

30.

Lee

W. B.

; Jacobs

D. S.

; Musch

D. C.

; Kaufman

S. C.

; Reinhart

W. J.

; Shtein

R. M.

Descemet's stripping endothelial keratoplasty: Safety and outcomes: A report by the American Academy of Ophthalmology. Ophthalmology 116(9): 1818–1830; 2009.

31.

Liu

; Zhou

; White

K. P.

RNA-seq differential expression studies: More sequence or more replication?

Bioinformatics 30(3): 301–304; 2014.

32.

Livak

K. J.

; Schmittgen

T. D.

Analysis of relative gene expression data using real-time quantitative PCR and the 2(-delta delta C(T)) method. Methods 25(4): 402–408; 2001.

33.

; Chen

; Liu

; Li

; Liu

; Zhou

; Wan

; Mou

Y. G.

; Wang

Enhanced survival in vitro of human corneal endothelial cells using mouse embryonic stem cell conditioned medium. Mol. Vis. 16: 611–622; 2010.

34.

Miyata

; Drake

; Osakabe

; Hosokawa

; Hwang

; Soya

; Oshika

; Amano

Effect of donor age on morphologic variation of cultured human corneal endothelial cells. Cornea 20(1): 59–63; 2001.

35.

Muhammad

; Peh

G. S.

; Adnan

; Law

J. B.

; Mehta

J. S.

; Yim

E. K.

Micro- and nano-topography to enhance proliferation and sustain functional markers of donor-derived primary human corneal endothelial cells. Acta Biomater. 19: 138–148; 2015.

36.

Nakahara

; Okumura

; Kay

E. P.

; Hagiya

; Imagawa

; Hosoda

; Kinoshita

; Koizumi

Corneal endothelial expansion promoted by human bone marrow mesenchymal stem cell-derived conditioned medium. PLoS One 8(7): e69009; 2013.

37.

Nayak

S. K.

; Binder

P. S.

The growth of endothelium from human corneal rims in tissue culture. Invest. Ophthalmol. Vis. Sci. 25(10): 1213–1216; 1984.

38.

Numata

; Okumura

; Nakahara

; Ueno

; Kinoshita

; Kanematsu

; Kanemura

; Sasai

; Koizumi

Cultivation of corneal endothelial cells on a pericellular matrix prepared from human decidua-derived mesenchymal cells. PLoS One 9(2): e88169; 2014.

39.

Okumura

; Kay

E. P.

; Nakahara

; Hamuro

; Kinoshita

; Koizumi

Inhibition of TGF-beta signaling enables human corneal endothelial cell expansion in vitro for use in regenerative medicine. PLoS One 8(2): e58000; 2013.

40.

Palchesko

R. N.

; Szymanski

J. M.

; Sahu

; Feinberg

A. W.

Shrink wrapping cells in a defined extracellular matrix to modulate the chemo-mechanical microenvironment. Cell. Mol. Bioeng. 7(3): 355–368; 2014.

41.

Peh

G. S.

; Chng

; Ang

H. P.

; Cheng

T. Y.

; Adnan

; Seah

X. Y.

; George

B. L.

; Toh

K. P.

; Tan

D. T.

; Yam

G. H.

; Colman

; Mehta

J. S.

Propagation of human corneal endothelial cells: A novel dual media approach. Cell Transplant. 24(2): 287–304; 2015.

42.

Peh

G. S.

; Toh

K. P.

; Ang

H. P.

; Seah

X. Y.

; George

B. L.

; Mehta

J. S.

Optimization of human corneal endothelial cell culture: Density dependency of successful cultures in vitro. BMC Res. Notes 6: 176; 2013.

43.

Peh

G. S.

; Toh

K. P.

; Wu

F. Y.

; Tan

D. T.

; Mehta

J. S.

Cultivation of human corneal endothelial cells isolated from paired donor corneas. PLoS One 6(12): e28310; 2011.

44.

Peh

G. S. L.

; Beuerman

R. W.

; Colman

; Tan

D. T.

; Mehta

J. S.

Human corneal endothelial cell expansion for corneal endothelium transplantation: An overview. Transplantation 91(8): 811–819; 2011.

45.

Price

M. O.

; Giebel

A. W.

; Fairchild

K. M.

; Price

F. W.

Jr. Descemet's membrane endothelial keratoplasty: Prospective multicenter study of visual and refractive outcomes and endothelial survival. Ophthalmology 116(12): 2361–2368; 2009.

46.

Schmedt

; Chen

; Nguyen

T. T.

; Li

; Bonanno

J. A.

; Jurkunas

U. V.

Telomerase immortalization of human corneal endothelial cells yields functional hexagonal monolayers. PLoS One 7(12): e51427; 2012.

47.

Senoo

; Joyce

N. C.

Cell cycle kinetics in corneal endothelium from old and young donors. Invest. Ophthalmol. Vis. Sci. 41(3): 660–667; 2000.

48.

Spandidos

; Wang

; Dragnev

; Thurber

; Seed

A comprehensive collection of experimentally validated primers for polymerase chain reaction quantitation of murine transcript abundance. BMC Genomics 9: 633; 2008.

49.

Spandidos

; Wang

; Seed

PrimerBank: A resource of human and mouse PCR primer pairs for gene expression detection and quantification. Nucleic Acids Res. 38(Database issue): D792–799; 2010.

50.

Sugita

; Usui

; Horie

; Futagami

; Yamada

; Ma

; Kezuka

; Hamada

; Usui

; Mochizuki

; Yamagami

Human corneal endothelial cells expressing programmed death-ligand 1 (PD-L1) suppress PD-1+ T helper 1 cells by a contact-dependent mechanism. Invest. Ophthalmol. Vis. Sci. 50(1): 263–272; 2009.

51.

Teo

B. K.

; Goh

K. J.

; Ng

Z. J.

; Koo

; Yim

E. K.

Functional reconstruction of corneal endothelium using nanotopography for tissue-engineering applications. Acta Biomater. 8(8): 2941–2952; 2012.

52.

Valtink

; Gruschwitz

; Funk

R. H. W.

; Engelmann

Two clonal cell lines of immortalized human corneal endothelial cells show either differentiated or precursor cell characteristics. Cells Tissues Organs 187(4): 286–294; 2008.

53.

Vianna

L. M.

; Kallay

; Toyono

; Belfort

Jr. ; Holiman

J. D.

; Jun

A. S.

Use of human serum for human corneal endothelial cell culture. Br. J. Ophthalmol.; 2014.

54.

Wang

; Seed

A PCR primer bank for quantitative gene expression analysis. Nucleic Acids Res. 31(24): e154; 2003.

55.

Weiss

J. S.

; Moller

H. U.

; Lisch

; Kinoshita

; Aldave

A. J.

; Belin

M. W.

; Kivela

; Busin

; Munier

F. L.

; Seitz

; Sutphin

; Bredrup

; Mannis

M. J.

; Rapuano

C. J.

; Van Rij

; Kim

E. K.

; Klintworth

G. K.

The IC3D classification of the corneal dystrophies. Cornea 27(10): S1–S42; 2008.

56.

Yee

R. W.

; Edelhauser

H. F.

; Stern

M. E.

Specular microscopy of vertebrate corneal endothelium: A comparative study. Exp. Eye Res. 44(5): 703–714; 1987.

57.

Yoshida

; Oshikata-Miyazaki

; Yokoo

; Yamagami

; Takezawa

; Amano

Development and evaluation of porcine atelocollagen vitrigel membrane with a spherical curve and transplantable artificial corneal endothelial grafts. Invest. Ophthalmol. Vis. Sci. 55(8): 4975–4981; 2014.

58.

Yoshihara

; Ohmiya

; Hara

; Kawasaki

; Consortium

; Hayashizaki

; Itoh

; Kawaji

; Tsujikawa

; Nishida

Discovery of molecular markers to discriminate corneal endothelial cells in the human body. PLoS One 10(3): e0117581; 2015.

59.

Yue

B. Y.

; Sugar

; Gilboy

J. E.

; Elvart

J. L.

Growth of human corneal endothelial cells in culture. Invest. Ophthalmol. Vis. Sci. 30(2): 248–253; 1989.

60.

Zavala

; Lopez Jaime

G. R.

; Rodriguez Barrientos

C. A.

; Valdez-Garcia

Corneal endothelium: Developmental strategies for regeneration. Eye (Lond.) 27(5): 579–588; 2013.